Guest Post by Willis Eschenbach
In my usual peripatetic wandering around the web, I came across an interesting paper called “Millennial- and orbital-scale changes in the East Asian monsoon over the past 224,000 years”, in Nature Magazine (subscription required), 28 Feb. 2008 , with Supplementary Online Information.
The paper uses “speleothems” to estimate past climate conditions. Speleothems are secondary mineral deposits formed in caves. Stalactites and stalgmites are speleothems, and they come in a wide variety of sizes and shapes. Here’s a photo of some speleothems:
Figure 1. Speleothems in a New Zealand Cave.
What can we learn from the speleothems?
The authors used the speleothem data from two caves in China to investigate the climate changes over the last two glacial periods, a quarter million years or so. Being more interested in the recent past, and noticing that one of the datasets extended up to the year 1490, I decided to see what speleothems could tell us about the temperature changes in more recent times. So I got a large group of speleothem records from the NOAA Paleoclimatology web site.
I wasn’t interested in what happened thousands and thousands of years ago, so I got all of the long records that covered all or part of the period from the end of the last ice age to the present. This gave me 20 records.
The speleothems give us a record of what is called the “delta oxygen 18” (∂18O) value. This value is related to the temperature. The paper does not give the associated temperature values, so I converted them using the relationship described here as:
This is based on the average d[delta]18O/dT relation in modern precipitation (~0.6‰ °C-1), and the water-calcite fractionation that accompanies speleothem deposition (~-0.24‰ °C-1).
Decoded, this means that the change in temperature is equal to the change in ∂18O divided by (0.6 – 0.24), or ∂18O/0.36. Using that relationship, I calculated the temperatures from the various speleothems, and graphed them all with no further adjustment.
Figure 2. Raw data from 20 speleothem records. All of them have been converted from ∂18O using the relationship Temperature = ∂18O/-0.36. Black line is a 200-point Gaussian average. Different records are different colors.
While this was interesting, it appeared to me that the various records were likely not vertically aligned quite properly. After all, there is no a priori reason to think that they would all fit together, since they were simple anomalies (data minus average of that data) over different time periods.
So how to adjust them? There are several methods that are used to make this kind of adjustment to temperature anomalies for the global temperature records. GISS takes an average of two records in the area where they overlap, and adjusts on that basis. That was possible here, but seemed inaccurate. GHCN, on the other hand, uses one type of “first difference” method. However, their method requires that all of the datasets be on the same basis (annual, monthly, etc.), where in this case the measurements are at various random times that differ between datasets.
After some thought, I realized that I could use “first differences” in another way. The “first difference” is a new dataset that is made by calculating the difference between successive datasets. For example, if the dataset is {1, 2, 4, 8, 10}, then the first difference of that dataset is {(2-1), (4-2), (8-4), (10-8)}, or {1, 2, 4, 2}. This represents the differences between the points in the original dataset.
I realized that the standard deviation of the first difference is a measure of how well the various datasets fit together. (Standard deviation, “SD”, is a measure of how scattered the data is.)
So to adjust them, I first combined all of the 20 speleothem datasets into one single large dataset. Then I took the first difference of that single dataset. I measured the SD of the first difference data.
Then I adjusted each of the individual speleothem records by moving it slightly upwards and downwards, and used the increase or decrease of the SD to indicate which way it should be moved. I repeated this until the match was not improved by further testing and moving of the individual datasets. The result is shown in Figure 3.
Figure 3. Adjusted data from the same 20 speleothem records. All of them have been adjusted vertically to give the best fit. Black line is a 200-point Gaussian average.
This has improved the accuracy of the reconstruction. This is shown by the greater vertical range of the Gaussian average line.
So, what does all this mean? Heck, I don’t know, I’m investigating, not drawing conclusions. A few comments, in no particular order:
• As is shown in the Greenland ice core records, we are currently at the cold end of the Holocene (the current interglacial).
• Recent phenomena (Roman Warm Period, Medieval Warm Period, Current Warm Period) are scarcely visible at this scale. So much for the “uprecedented” nature of the recent rise.
• The polar bears are not in any danger from the recent rise.
• What’s up with the big jump and drop about 12000 years ago? I have not seen that in the ice core records, but it is present in these speleothem records from around the planet. [Update] A number of people have pointed out that this is almost certainly the “Younger Dryas” event. I hadn’t noticed it in the Vostok record, but a closeup of that record shows it.
• The amount of the temperature change depends on the coefficient used to translate from d18O to temperature. So the numbers are likely in the right range, but may be somewhat too large or too small.
Anyhow, that’s my thoughts about what I’ve found out, I welcome yours. I continue with the investigation. It strikes me that I may be able to adjust the conversion factor (d18O/T) to see if that improves the fit of the data … should be interesting. Onwards …
DATA:
The caves used in this study were:
Cave, Location
Borneo_sch01, Borneo
Borneo_sch02, Borneo
Buckeye, Central US
Chilibrillo, Panama
Cold_Air, South Africa
Crystal, Midwest USA
Dayu, Central China
Dongge, Eastern China
Dongge04, Eastern China
Dongge05a, Eastern China
Heshang, Central China
Liang_Luar, Indonesia
Lianhua, Southern China
Lynds, Tasmania
Mystery, Midwest USA
Sanbao08, Central China
Sanbao10, Central China
Soreq_Bar, Israel
Spannagel, Austria
Venado, Costa Rica
In two cases, where there were several speleothem records from the same cave analysed by the same investigators, I have combined them into a single longer record. Data from different studies of the same cave have a year (e.g. “08”,”10″) appended to the name.
I have posted the data I used, along with the R file that I wrote to analyze the data, as a zip file here. Enjoy!
sure sure. you’re right. golf ball bathroom. got it.
Now we can stop complaining about tone.
you say: “So you are saying that the record I showed is a precipitation record that also shows temperature? It’s very hard to try to dig out your meaning.”
Did you read the wang paper? That should clear up what I am saying.
William Roberts says:
June 3, 2010 at 4:02 pm
No, the Wang paper can’t clear up what you are saying. Only you can clear up what you are saying. So let me try again.
Are you saying that the record I have shown is a precipitation record that also shows temperature?
uhhhh, yes it can. I don’t want to have to tell you the play-by-play of the whole paper when you can just read the paper yourself. Sorry, trying to be scientific and not lecture you about it. Let me know when you’ve actually read it so we can discuss the science more.
Oh, and about your in-depth analysis of the Lyman paper: so you’re judging a book by its cover (or title as it were) now? Good call. And given that each year in the paper consists of hundreds, if not thousands of data points from buoys and floats, your 16 degrees of freedom and resulting p-value mean squa-doosh. So the trend, which you confirm, is “robust”. …but of course you would’ve known that if you actually read the paper as opposed to conducting your rigorous scientific analysis without even reading it.
I posted a reply this morning. Will it show up later? Or are you censoring posts?
REPLY: Or maybe people have lives and we just finished getting the kids ready for school? – Anthony
nevermind. It just showed up. Weird. sorry for multiple posts.
William Roberts says:
June 3, 2010 at 4:02 pm
I replied:
William Roberts says:
June 4, 2010 at 6:48 am
OK, I read the Wang paper. Funny, I didn’t see your name in it anywhere, maybe I missed it. Unfortunately, they say what they said … and you say what you said. But Wang et al. don’t say a single word about what you said.
So if you could answer my question, viz:
Because no matter how I read Wang, I don’t see the answer in there as to how you view the data I show in this thread.
First off – wow this post is already on page 3 of this blog. you guys really churn them out here. …I doubt anyone is actually even reading this thread anymore as it is already buried. So I guess it’s just you and me. Too bad b/c finally people may understand the mistakes you have made.
Second – nice double talk. Way to avoid addressing my comments by repeating the question.
third – yes, I DID in fact answer your questions. The change in precipitation from glacial times to present occurred b/c the earth warmed up. Warmer earth, more rain in many places. Over the course of the Holocene, there is a phase-shift of 180 degrees between the northern and southern hemispheres. That’s why the China speleothem and the brazilian speleothem records have opposite trends over the Holocene. As is VERY evident in the paper. And is what I said here:
“The earth was deglaciation. Continental ice caps were melting. That little speed bump caused worldwide changes – including, but not limited to, temperature and precipitation.
Alternatively, over the course of the Holocene, insolation changes caused the see-saw pattern in hemispheric precipitation. And I DID give you a ref. You should’ve read it before you asked your question. So reading comprehension again. Read the Wang ref.”
Therefore, you are incorrect again. We both said the same things. ..also, how many times do I have to say it???
Fourth – why would you see my name in the wang paper? I didn’t have anything to do with it. It is part of the answer to why your post is incorrect. That’s why I shared it with you. …now who is being uncivil???
William Roberts says:
June 7, 2010 at 9:08 am
Man, you are tap dancing around answering my question like Gene Kelly on steroids. Lets go over it again, I’ll make it simple for you:
[ ] Yes
[ ] No
Somehow, the Wang paper doesn’t answer that. Somehow, you don’t want to answer that.
But I’m a patient man. Answer it, and we can move on to the other mysterious “mistakes” that you claim I’ve made, but that you don’t specify. You said my big mistake was saying that what I showed reflected temperature. So are you standing by that, or not? Let’s get past that one, and then we can move on to the others. Perhaps the “mistakes” are in this statement of mine:
Finally, you ask, “… now who is being uncivil?”
You were very uncivil when you started out. You have falsely accused us of censoring your posts. You have said that my actions might be a “devious ploy”. You have not commented on my analysis of the statistical significance of the Lyman paper, but you have roundly abused me for it nonetheless. If you think my analysis of the statistical significance is wrong, perhaps you might deign to tell us why, rather than attacking me. You have accused folks here of “blindly accepting” what I say, when in fact I have faced a host of questions about what I’ve said.
So in answer to your question about who is being uncivil … well, that would be you.
If I’m Gene Kelly, then you are Cyd Charisse.
What I’m saying is this: speleothems record a combination of temperature and precipitation changes. If you have a way of determining the change in temperature via other means, say a nearby marine sediment core – then you have an independent constraint on temperature. You can convert that temperature to d18O of calcite using a thermodynamic equation. The resultant change in calcite d18O cannot be due to temperature. It is due to amount or the source of precipitation.
clear enough??? probably not, so go ahead and fire away.
And I did comment on your analysis of the Lyman paper. I said you didn’t properly count the degrees of freedom for your p-value. reading comprehension.
As to who is uncivil??? Both of us are guilty. I can live with it. Can you???
William Roberts says:
June 9, 2010 at 1:41 pm
Sorry, I missed it. You had said:
I like “squa-doosh”, great term. However, the fact that each year consists of a large number of data points only reduces the standard error of each data point. It does not affect the resulting “p-value”. The authors know this, which is why they calculate the p-value using the number of annual data points, not the thousands of individual data points.
Certainly. My point was quite simple. If you enter a discussion by abusing people and calling them stupid, you’ll get it back in your face. I was the one who put your incivility back in your face, and I make no apologies for that. You may be correct in your claims that we are all dumb and that I haven’t a clue … but it’s damn poor tactics to walk in the door saying so if you actually want to accomplish something …
SO- back to your original post where you put absurdly large temperature variations for cave records in the tropics and sub-tropics: do you plan on re-posting with an update? …Or just leave your readers with false information.
This post is already on page 4 of this blog, so you guys aren’t afraid to crank out a post.
Question regarding the p-value: if they calculated 6 month averages instead of annual using the thousand of data points and then you did your trend significance — would it be significant then???
How about 4 month averages (ie. seasonal)???
William Roberts says:
June 14, 2010 at 10:17 am
You’re not following the thread. I did an update addressed directly to you some time ago.
William Roberts says:
June 16, 2010 at 7:35 am
Not sure what your point is here. Statistical significance is definitely affected by your “binning” decisions (averaging monthly, quarterly, annually). I can only comment on the binning that they actually used.